Re: [PATCH v14 8/9] KVM: x86: virtualize cpuid faulting

[Andy Lutomirski]

On Fri, Jul 27, 2018 at 2:30 PM, Jim Mattson <jmattson@xxxxxxxxxx> wrote:
> On Fri, Jul 27, 2018 at 2:05 PM, Andy Lutomirski <luto@xxxxxxxxxx> wrote:
>> Does KVM *have* a concept of "maximum non-turbo frequency" of the
>> guest that it would make sense to expose here?  If so, presumably the
>> right solution is to expose it.
>
> KVM has the concept of a guest's invariant TSC frequency. The Maximum
> Non-Turbo Ratio is just some fraction of that. Sadly, the fraction is
> 100 MHz, 133.33MHz, or the "scalable bus frequency" from some other
> MSR, depending on microarchitecture.

Which is problematic, unless KVM wants to start deciding what the base
clock is.  There's MSR_FSB_FREQ, which is supported on Atom only,
IIRC.

I really wish Intel would get its act together.